Backups that actually restore.

// The RAID trap

RAID is redundancy, not backup.

"It's on the NAS, it's RAID, we're covered" is the most expensive sentence in small-business IT, so here's the reality from my own rebuild log. When a drive in a ten-drive array failed and was replaced, the rebuild estimate came in at close to a week at the default rebuild rate. For that entire week the array runs degraded: slower for everyone using it, with the redundancy you bought it for already spent. One more failure inside that window and everything on the array is gone. That is what RAID actually buys you. It keeps you running through a single drive failure, and nothing more. It does nothing against deletion, ransomware, theft or fire, and rebuild windows only get longer as drives get bigger.

So the plan needs two parts. A real backup that lives somewhere the array's problems can't reach, and a rebuild plan written before you need it: how long the rebuild takes, whether to raise the rebuild rate at the cost of daytime performance, and a fresh backup taken the moment the array goes degraded, before the rebuild starts. If your data only exists on one RAID box, you don't have a backup. You have one failure between you and nothing.

// The rule

The 3-2-1 rule, applied to a Sydney office.

Three copies of your data, on two different types of media, with one copy off-site. It has survived every kind of disaster because it is that simple. Here is what it looks like for a typical office: a handful of workstations, one server or NAS, and a Microsoft 365 tenancy.

• Copy one is the live data. The files on the server, the mailboxes in Microsoft 365, the documents on each laptop. This is the copy RAID protects, and RAID copies a deletion or a ransomware encryption to every disk just as faithfully as your good work.
• Copy two is local. A backup on separate hardware on-site, on a different kind of media from the original, so one fault or one attack cannot take both. It is the copy you restore from most days, because it is fast: a deleted folder, a corrupted file, a laptop that will not boot. Parramatta and Western Sydney on-site: see IT support Parramatta.
• Copy three is off-site. Kept somewhere else entirely, so a fire, theft, flood or ransomware at your premises cannot reach every copy. This is the copy that beats ransomware, because a copy the attacker cannot reach is a copy the attacker cannot encrypt.
• Workstations count. The spreadsheet that runs the business is often on one person's desktop, not the server. Workstation backup is $33 per device per month.
• The server gets both. Server backup, local plus off-site, is from $110 per server per month.
• The cloud tenancy counts too. Microsoft 365 is copy one for your email and SharePoint files, not a backup of them. Microsoft 365 backup is $5.50 per user per month.

The long version, including what RAID is genuinely good at, is in RAID is not a backup: the 3-2-1 rule explained. Whether you need one server, a cluster or the cloud in front of those backups is a separate question, answered in how much redundancy a small business needs.

// Off-site

Cloud backup, and where the off-site copy actually lives.

"It's in the cloud" is only a backup answer if you know which cloud, who runs it, and whether you can get the data back. Most providers rent space on someone else's computers. I run my own data centre on Australian soil, and I work with Microsoft and the big clouds too, so the off-site copy goes where it fits rather than where the margin is.

• On infrastructure the same person runs. The off-site copy can live in our own self-hosted data centre, on Australian infrastructure with a 99.98% uptime target, watched by the same monitoring that watches everything else I run. When you call about it, you are talking to the bloke who runs the hardware.
• Or on a big cloud where that fits. Some workloads belong on Microsoft's cloud or Amazon's. Because the setup works either way, there is no reason to push you toward one or away from the other.
• Out of ransomware's reach. The off-site copy is not a mapped drive that is always connected and writable. Ransomware encrypts everything it can reach, including a backup drive left permanently plugged in. A copy the attacker cannot reach is the whole point.
• Restores, not just backups. Backup software reporting success tells you the job ran, not that the data is usable. Restores are tested by actually recovering files, on a schedule, and the date goes in the plan.
• Your data stays yours. No lock-in is a design rule: you own your accounts and your data, and every setup has a documented exit path.

Where the rest of your systems should live, hosted, in-house or hybrid, is the cloud solutions page. The data centre itself is on hosting and data centre.

// The plan

Disaster recovery plan: what is in the document.

A disaster recovery plan is the set of decisions that are hard to make well at 7am with your systems down and staff ringing. Making them now costs an hour. The documented and tested plan is from $1,650, and it contains these sections.

• Critical business functions. What must keep running, and how long each can realistically stop. This is where the recovery time objectives come from: an agreed number per system, written down, not a slogan.
• Systems and data inventory. Where the data actually lives, who provides it, where the backup is, and how long a restore really takes. If the backup column says "cloud", treat that as an unanswered question rather than a tick.
• Disruption scenarios, each with first-hour actions. Internet down, power out, premises inaccessible, ransomware, email compromise and invoice fraud, key person unavailable, hardware failure, supplier failure.
• Key contacts, including personal addresses. If the outage is your email, you cannot use work addresses to coordinate.
• Access and credentials. Where things are kept, never the passwords themselves, and who at the business holds the domain registration and the Microsoft 365 global administrator login.
• A testing log. An untested plan is a document, not a capability. The annual DR test, from $880 per year, restores real systems on purpose and gives you a written report with the date in the last-verified-restore column.
• Who holds it. You do. The first-hour checklist is one printable page that applies whatever went wrong, because a plan that only exists on the file server is no use when the file server is the thing that is down.

If you want to start yourself, the business continuity plan template is free, in Word, with no email address required. Ten sections, ten tables, about an hour to fill in. Most people finish it with two or three uncomfortable blanks, and those blanks are the conversation.

// Microsoft 365

Microsoft 365 backup: what Microsoft does not keep.

Microsoft protects the platform. It keeps the service running and its own hardware redundant. It does not promise to hand back the file a staff member deleted last month, the mailbox of someone who left, or the SharePoint library ransomware encrypted and then cheerfully synced.

• Departed staff. When a licence is removed, the mailbox is retained for 30 days, and then it is gone. Converting a leaver to a shared mailbox keeps the mail for nothing, but that is retention, not a backup you can restore from a point in time.
• Deleted and overwritten files. Deleted mail, ransomware and broken sync jobs travel to the cloud as happily as everything else. The sync client uploads the encrypted versions just as faithfully as the good ones.
• Old mail on local machines. Mail in local data files on individual computers lives nowhere else and is not part of any server or tenancy backup. It has to be found and brought in.
• What the backup does. A separate copy of mailboxes and SharePoint files, kept apart from the tenancy, that you can go back to from an earlier point in time. Microsoft 365 backup is $5.50 per user per month.
• Restore-tested like everything else. Same rule as the server: a backup nobody has restored from is a rumour.

Tenancy setup, the security baseline and licences at RRP are on the cloud and Microsoft 365 page. Which addresses do not need a paid licence at all is in shared mailbox in Microsoft 365, and the catches in moving mail across are in email migration to Microsoft 365.

// Emergency

Emergency recovery for non-clients.

You do not have to be a client. If a drive has failed, an array has corrupted or camera footage has to come back, emergency recovery is $220 per hour, and the work is done in-house rather than shipped to a lab. That is unusual capability for a provider this size, and it has saved real businesses real money.

• Failed drives. Workstation and server drives that have stopped being detected, or that mount and show an empty disk.
• Corrupted arrays. A RAID that lost a second disk inside the rebuild window, which is the one failure RAID cannot survive, or a NAS that will no longer assemble its array.
• Camera footage. Recorders nobody has the password to, and footage that has to be pulled off a system that was never meant to give it up.
• Stop using the device. The most important thing you can do. Continuing to write to a failing drive is what turns a recoverable situation into an unrecoverable one. Ring before trying DIY recovery software.
• Honest odds first. If recovery is likely to cost more than the data is worth, you will be told so rather than quietly billed for it.

Once the data is back, the conversation turns to why there was no second copy. That is the free consult, and it is the reason this section sits on the backup page.

// Indicative pricing

What it roughly costs.

Real numbers, because "contact us for pricing" is code for "brace yourself." These are honest ballparks, your fixed quote comes after a free consult.

Indicative pricing ex GST at market rates +10%. If you're not sure your current backups work, that's the free consult, bring your worst fears.

Every small business IT support plan we run includes backup checks with tested restores. For the other controls that stop ransomware, read our guide to cyber security for small business in Sydney.

// Questions

Backup, your questions answered.

// Pairs well with